Increase in lifetime adverse drug reactions, service utilization, and disease severity among patients who will start COX-2 specific inhibitors: quantitative assessment of channeling bias and confounding by indication in 6689 patients with rheumatoid arthritis and osteoarthritis.

OBJECTIVE

Nonrandom assignment of therapy in observational studies and clinical practice can be accompanied by channeling bias and confounding by indication. This in turn can lead to unreliable conclusions about treatment effectiveness. Although widely acknowledged as important, no studies in rheumatology have measured the extent of these biases. We identified variables contributing to confounding and investigated the strength of the confounding effect. Analytical methods (propensity scores) are available to mitigate the effect of nonrandom assignment if the full extent of confounding can be understood.

METHODS

A population of 6637 patients with rheumatoid arthritis (RA) and osteoarthritis (OA) from the practices of 433 US rheumatologists completed 2 sets of detailed questionnaires concerning (1) the last 6 months in 1998 and (2) the first 6 months of 1999, generally prior to and after the release of celecoxib and rofecoxib. Patients who received the COX-2 specific inhibitors in period 2 were identified (n = 1517), and their characteristics were compared to the 5120 who did not start a new COX-2 specific inhibitor during Period 1.

RESULTS

Patients starting a new COX-2 specific inhibitor had a greater lifetime history of adverse reactions of all kinds, but particularly gastrointestinal adverse drug reactions. They also had more severe scores for pain, functional disability, fatigue, helplessness, and global severity, and they used more inpatient and outpatients services than patients who would not switch to COX-2 specific inhibitors.

CONCLUSIONS

Confounding by indication and channeling bias result in an overall increase in severity of about 25% for the above measures. Observational studies should account for these biases by a broadly defined propensity score that includes the variables identified in this report. These observations are germane to observational studies of disease modifying antirheumatic drugs and biologics, as well, and suggest the need for careful control of confounders when assessing treatment effects in rheumatic disease observational studies.